Discovering the links between real-world activities and previous course contents: the potential of information retrieval using large language models

Abstract

In experiential learning involving real-world activities, such as fieldwork and pre-service training, transferring knowledge into practice is essential. While reflection is a critical component of this process, it is challenging to review how previously learned knowledge has been utilized. To address this issue, this study connected student descriptions of real-world activities with relevant course contents using information retrieval techniques and large language models (LLMs). The validity of linking was evaluated for one approach without LLM and three approaches that employ LLMs differently. These approaches were applied to a dataset collected from a university course in Japan. There were conditions for the inclusion or exclusion of supplemental information. The results indicated the supremacy of LLM-featured approaches without supplemental information. However, we found that these performances have not yet been stable. The findings and discussions shed light on the potential of the LLM-featured retrieval approaches for data-enhanced reflection across in-class knowledge acquisition and real-world knowledge applications.